Why does the pump make any difference?

Because it changes the pressure in the pipework, and pressure is what pushes water out through a fault. The return pipework is pressurised only while the pump runs, so a fault there leaks hard when it is on and barely at all when it is off. Nothing about the pool itself changes.

Why would a pool lose more with the pump OFF?

Because the suction pipework is the opposite case. While the pump runs, that side is under negative pressure, so a fault there pulls air in rather than pushing water out. Switch the pump off and the negative pressure goes, leaving the water in the pipe held only by its own head, which then escapes through the same fault.

Are there other signs of a suction-side fault?

Air is the giveaway. If a fault on the suction side is drawing air in while the pump runs, that air has to leave somewhere, and it leaves through the returns. Bubbles coming back into the pool while the pump is running are consistent with air being drawn in upstream of it.

Do I have to run the test twice?

To get a location, yes. One test tells you how much you are losing and two tell you roughly where from, and the second test costs one more day. It is the best value day of work available on a leaking pool, because it decides whether anybody needs to dig.

How different do the two results need to be?

Meaningfully rather than marginally. Two measurements a day apart will never match exactly, because the weather changes and a millimetre of measurement error on a 32 m² pool is 32 litres. A difference of a few per cent is noise. A difference you would describe as obvious is a signal.

Pool losing water only when the pump runs

A pool that loses water faster with the pump running has a leak on the pressure side of the circulation, because that pipework is only under pressure while the pump is on. A pool that loses faster with the pump off points at the suction side instead. A pool that loses at the same rate either way is losing it through the shell, the liner or a fixed fitting.

Last reviewed 14 August 2026

The short answer

Run the bucket test twice, once with the pump running for the whole 24 hours and once with it off for the whole 24 hours, and compare the two net losses.

Result	Where the loss is	Because
Faster with the pump ON	Pressure side of the circulation: returns, and the pipework after the pump.	That pipework is pressurised only while the pump runs, and pressure is what pushes water out through a fault.
Faster with the pump OFF	Suction side: skimmer lines, main drain line, and the pipework before the pump.	While the pump runs, that side is under negative pressure and draws air in rather than pushing water out. With the pump off, the water escapes under its own head.
The same either way	The shell, the liner, or a fixed fitting.	A structural loss is not connected to the circulation, so it does not care what the pump is doing.
No net loss in either test	Nowhere. There is no leak.	Both results were evaporation, which the bucket accounted for.

What the comparison tells you

This is the single most useful thing the bucket test produces, and it costs one extra day. It is the difference between knowing you are losing water and knowing which half of the system to look at.

How this works in practice

The mechanism is pressure, and the two sides of a pool circulation system are under opposite pressures while the pump runs.

The pressure side runs from the pump outlet, through the filter and heater, to the returns. While the pump is on it is above atmospheric pressure, so any fault pushes water out. Switch off and it depressurises, and the leak slows or stops.
The suction side runs from the skimmers and main drain to the pump inlet. While the pump is on it is below atmospheric pressure, so any fault pulls air in. Switch off and the vacuum goes, and the water in the pipe leaks out under its own head instead.
The shell and the liner are neither. They hold water because they are watertight, not because of anything the pump does, so a fault there behaves identically in both tests.

That gives a free confirming observation for the suction case, which follows from the mechanism rather than needing to be looked up. If a suction-side fault is drawing air in while the pump runs, that air passes through the pump and leaves through the returns. Bubbles coming back into the pool with the pump running are consistent with exactly that.

The structural case has its own follow-up, because where the level stops falling narrows it further. whether the loss is in the liner or the structure covers what a self-stopping loss means.

What changes the answer

Three things, and the first is why a marginal difference should not be read as a signal.

Measurement error. A millimetre on a 32 m² pool is 32 litres, and two tests on two different days will never match exactly. Treat a few per cent as noise and an obvious difference as a result.
The weather. The bucket corrects for evaporation on each day separately, which is exactly why this comparison works at all, but a very windy day and a still day are still different conditions and a bigger evaporation correction carries a bigger error with it.
Anything that adds water. An auto-fill device makes both tests meaningless, because the pool is being topped up while you measure it. Isolate it before either test.

If the two results are close and both show a loss, that is a structural answer rather than an inconclusive one. Same either way is a diagnosis, not a failure to find a difference.

Assumptions and sources

Litres in a cubic metre

Exact by definition

SI definition. The litre is defined as one cubic decimetre, so one cubic metre is exactly 1,000 litres.

Millimetres in a metre

Exact by definition

SI prefix definition. Milli is a factor of one thousandth.

Difference that counts as pump dependent

Illustrative assumption, editable above

Illustrative assumption. Two bucket tests run on different days differ anyway, because wind, air temperature and sunshine differ. This is the difference between the two daily losses, as a fraction of the larger, below which the two tests are treated as the same result rather than as evidence of a pressure-side leak.

Nothing on this page rests on a constant. That the pressure side is above atmospheric pressure while the pump runs and the suction side below it is how a circulation system works, and the consequences follow from it rather than from a measurement.

The threshold this site uses for deciding whether two results differ meaningfully is a practical one rather than a measured boundary, and like every other band on this site it renders as an adjustable figure rather than a hidden constant.

What to do next

Three things, and they follow directly from which of the four results you got.

Pressure side: the return pipework and its fittings are where to look, and a pressure test on that section is the standard next step. It is also the case most likely to be a buried pipe rather than something visible.
Suction side: check the skimmer throats and the pipework before the pump first, and watch the returns for bubbles with the pump running as a free confirmation.
Structural: note the level at which the loss stops, if it stops. That single observation narrows a structural loss down more than another day of testing will.

In every case, confirm the volume at the meter as well if the pool is on a metered supply. confirm the rate of loss at the meter catches the case an auto-fill device would otherwise hide completely.

Common questions

Why does the pump make any difference?: Because it changes the pressure in the pipework, and pressure is what pushes water out through a fault. The return pipework is pressurised only while the pump runs, so a fault there leaks hard when it is on and barely at all when it is off. Nothing about the pool itself changes.
Why would a pool lose more with the pump OFF?: Because the suction pipework is the opposite case. While the pump runs, that side is under negative pressure, so a fault there pulls air in rather than pushing water out. Switch the pump off and the negative pressure goes, leaving the water in the pipe held only by its own head, which then escapes through the same fault.
Are there other signs of a suction-side fault?: Air is the giveaway. If a fault on the suction side is drawing air in while the pump runs, that air has to leave somewhere, and it leaves through the returns. Bubbles coming back into the pool while the pump is running are consistent with air being drawn in upstream of it.
Do I have to run the test twice?: To get a location, yes. One test tells you how much you are losing and two tell you roughly where from, and the second test costs one more day. It is the best value day of work available on a leaking pool, because it decides whether anybody needs to dig.
How different do the two results need to be?: Meaningfully rather than marginally. Two measurements a day apart will never match exactly, because the weather changes and a millimetre of measurement error on a 32 m² pool is 32 litres. A difference of a few per cent is noise. A difference you would describe as obvious is a signal.